DK_en 2x04 - Bullshit Generators

Episode first aired on 6 March, 2023. Listen on Spreaker.com.

In this episode, I will fairly often use the word "bullshit". I want to stress that my use is entirely academic. In 2005, Harry Frankfurt, Professor Emeritus at Princeton, published a small but brilliant book. its title was "On Bullshit".

In the book, prof. Frankfurt made the fundamental distinction between lies and, you guessed it, bullshit. Although both types of discourse aim to convince, liars know they are lying, and do it on purpose. Bullshitters, on the other hand, have no interest whatsoever in the truth. A bullshitter's only motivation is to convince.

The concept has become very useful in an era where any marketing delusion can be fuelled and given an aura of scientific dignity by cherry-picked Big Data, and other academics have started using it, like Bergstrom and West in "Calling Bullshit", which is a blog, a book, and a university course.

Why am I mentioning this?

Well, openAI's chatGPT, Google's Bard, Microsoft's BingGPT, Facebook's already forgotten Galactica, and Large Language Models in general, can produce convincing text with no necessary connection to factual truth.

My view is that these systems are Bullshit generators, and I am fairly convinced that calling them such is a healthy antidote to overestimating and anthropomorphising Large Language Models, which is what the cultist propaganda around them would like to have us do.

A dear friend (hello Riccardo) told me the other day about a Facebook posting by some guy. Incidentally, my facebook profile has been dormant for years.
I do not know the author, but he is a Senior Researcher at the NATO defence college. So not the kind of person from whom you'd expect a catchy drivel, but 'social is social' is the modern version of 'opportunity makes the man a thief.'"
What did the guy say? This:

OpenAI was founded in 2015. GDPR was defined in 2016. Now OpenAI has chatGPT. What's does Europe have?

Apparently being a senior researcher for NATO does not automatically guarantee logical prowess. But never mind the finessing about false equivalences. This is one of those cool, non-committal social nonsense; badmouthing the GDPR, especially with empty arguments, is the couture version of "politics is for thieves".

However. There is something about this particular drivel, a populism just slightly subtler than usual, something profoundly idiotic and yet with an almost endearing surface.
I had to think about it for a while, and then I got there. It is not nonsense. It is propaganda.
It is the 'place in the sun' theory that Mussolini played long ago. If you have any high school history left, you may remember that, long before he entered the war on the side of Germany, at a time when he was particularly well liked overseas for how he had managed to wipe out unions, Mussolini founded the Empire, something that to this day brings a tear to the eye of those who always go out of their way to remind us that the founder of Fascism did good things too, after all.

Here is what Mussolini said on 2 October 1935:

"For many months the wheel of destiny, under the impulse of our calm determination, has been moving towards its goal: in these hours its pace is faster and unstoppable now! It is not only an army tending towards its goals, but it is a whole people of forty-four million souls, against whom the blackest of injustices is being attempted: that of taking away our place in the sun_"

Now, ninety years later, I find sublime the dichotomy between the operetta-like machismo of these speeches and the image of a place in the sun, which is so much retirees in Marbella. Yet it was with this speech that Mussolini announced the invasion of Ethiopia, i.e. the aggression against a sovereign country, in an attempt to expand Italy's colonial possessions after Eritrea, Libya and Somalia.

Why do I waste time talking about Mussolini? Because propaganda always works the same way, and in fact today the logic with which they want to sell us chatGPT and assorted bullshit generators is the same: everybody does it, why not us?

Now, I think that this line of reasoning should play no part in adult discussion. Outside of propaganda, the motivation to imitate someone can only be based on the fact that they are doing something worthwhile, worth following. In the case of colonialism, the answer was obviously no.
And in the current case of bullshit generators, it's also no.

Before fearing being left behind by chatGPT, one should ask oneself where a text generator incapable of distinguishing whether the claims it produces are true or false leads.

But why do I speak of propaganda? Because that is what it is all about. The day before the facebook post, none less than the Guardian came out with this headline: 'The UK needs its own BritGPT or it will face an uncertain future'. The piece was about a speech by British Telecom's Chief Data and Artificial Intelligence Officer to the House of Commons Science and Technology Committee; in the speech the following was said:

"We believe that there is a risk that we, in the UK, are losing ground to the big technology companies, and maybe even China, and falling behind... in the areas of cyber security, healthcare and so on. It's an arms race that has been going on for a long time, but has intensified in recent times'.

What automatic bullshit generation might have to do with cybersecurity and even healthcare, God knows. But the point, of course, is to sell the emergency.

The chorus of whiners has been joined by:

• Dame Wendy Hall, who chaired the UK government's enquiry into AI in 2017,
• by the director of the Open Data Institute,
• and even by the Tony Blair Institute, which wrote a report on purpose.

All agreed that not only do we need to invest, but we need to invest right there: in bullshit generators. Not in some obscure thing that looks like something out of an actual research centre, no, in that very thing that even the TV is talking about.

And it's not all.

Forty-eight hours later, the Wall Street Journal published an op-ed in what I would call a measured tone:

Headline: "ChatGPT announces an intellectual revolution"
Lead: "*Generative artificial intelligence represents a philosophical and practical challenge on a scale not seen since the beginning of the Enlightenment."
Authors: Henry Kissinger, Eric Schmidt and Daniel Huttenlocher

Henry Kissinger is the political champion of the worst imperialist capitalism we all know, Eric Schmidt is the former CEO of Google, and Huttenlocher is a bigwig at MIT. Two years ago, the three wrote a book together that also promises to be measured: 'The Age of AI and the Future of Humanity'. It may be a coincidence but the three musketeers come out in the WSJ one day after the Guardian article.

All swearing and perjuring that with chatGPT they have seen the light, that chatGPT is the way, the truth and the life. To me, it seems quite a stretch for a piece of software capable of generating the story of bears in space on command, or suggesting that an infant's diet be enriched by adding crushed pottery to mother's milk.

I miss the piece where someone proves that bullshit generators, which everyone calls by the market name of 'generative artificial intelligence' in order pretend to be speaking of serious stuff, really do have uses that live up to their promises. And I will continue to miss it, because it will never come.

So what's all the fuss about? My guess is that, commercially, you win by placing the products that pull, and Silicon Valley hasn't come up with something really new in almost forty years, since Apple introduced the world to graphical user interfaces with the Macintosh. Incidentally, that wasn't new stuff either, Jobs unashamedly stole Xerox PARC's ideas. He did it badly, but the leap in usability over character-based interfaces was such that everyone was happy with it.

If you put a 1987 Macintosh with System 4.1 and a MacBook with MacOS 13 side by side, what changes? For the user's life and work, I say. Take away all the decoration, 3D-icons, window animations: what's different?

The Web. The Web was the real revolution. Everything is the normal technical improvements.
And the Web was born in Europe.

In 1987 a Macintosh, then the most powerful personal computer on Earth, had

• a 32-bit Motorola 68000 clocked at 7, 83MHz,
  • it was called that because it had 68000 transistors
• 1Mb of RAM,
• 20MB of disk
• 1.4MB floppy

a MacBookPro today has

• an 8-core Apple M1 with 5nm technology and a 3.20GHz clock =(400x)
• 16 billion transistors (=235,294x)
• 16GB of RAM (=1600x)
• 500GB of disk (=25,000x)

And for the user they do the same thing, a hair faster but not always.
Silicon Valley has seen no real innovation for forty years.

Now you see that failing to innovate, the second choice is to sell old things that look new. Enter social networks. Facebook is the Web interface to a BBS, Twitter and assorted IMs are Web interfaces to IRC.

But even social networks are declining, so Big Tech to remain relevant needs something that sounds new, that people want, and above all that plays to Big Tech's strengths, something that requires enormous resources: enter the metaverse and AI. But the metaverse will go down in history as the Big Yawn of the 2020s.

I mean it's so revolutionary that my daughter only plays roblox or Minecraft a couple of times a month because she doesn't enjoy it anymore, and she's fourteen. So that leaves only AI, whatever that may be.

And even with AI, boredom flows powerfully. But then OpenAI comes up with a chatbot on steroids, the public finally finds some interest, if only to have a laugh, and the choir of the Church of Technology can start singing again.

chatGPT will replace journalists, lawyers, judges, change education, change online research, and oh my God, students now have chatGPT write their essays and nothing will ever be the same again, pinky swear.

The only job that chatGPT can really replace is the production of keyword-laden text with no constraints on reality, to the delight of content farms and marketers. And it is work that has never really had a reason to exist, except as a side effect of Google's choice to allow any 'SEO expert' to cheat with their search ranking.

But never mind. The business of Silicon Valley is not to innovate, it is to make money. ChatGPT seems to promise tons of money because, provided you sow the fear of falling behind it, no one will want to start from scratch or do anything different.

It will happen like with that pathetic GAIA-X thing. Strategic, European, and in the hands of Microsoft, Amazon and Google because only they can guarantee the performance no one has ever asked for in public information systems. But who else can guarantee hyperscaling? Finally, even the website of the Ministry of Health will be able to cope with traffic peaks of three orders of magnitude, should Taylor Swift or Måneskin want to tour hospitals.

After having convinced our rulers that no one in Europe can maintain a decent data centre, now Big Tech wants to sell us that there is no computer research if there is no Artificial Intelligence, and there is no Artificial Intelligence if not in bullshit generation.

And you can be sure that very soon throughout Europe the big research centres will join the whine, they too will want their place in the sun.

Only, those roundish things that chatGPT leaves along its path, you see, are not small suns.